Weaker version of Hall's condition












0












$begingroup$


In chapter 1 of Lubotzky's "Discrete Groups, Expanding Graphs and Invariant Measures", remark 1.1.2, one definition of expanders is the following: if $n, k in mathbb{N}$ and $c in mathbb{R}_{> 0}$, an $(n, k, c)$-expander is a $k$-regular bipartite graph with $n$ inputs and $n$ outputs such that for any set $A$ of at most $frac{n}{2}$ inputs we have $|partial A| geq (1 + c) |A|$, where $partial A = { v : d(v, A) = 1 }$.



The author goes on to say that any such graph admits a perfect matching by Hall's marriage lemma. Now I know that regular bipartite graphs always admit a perfect matching. However some other authors define expanders as graphs with degree bounded by $k$, instead of $k$-regular graphs. So I was wondering if the following is true:



Let $X$ be a bipartite graph with $n$ inputs and $n$ outputs and maximum degree $k$, such that for any subset $A$ of at most $frac{n}{2}$ vertices we have $|partial A| geq (1 + c)|A|$, for some fixed $c > 0$. Does $X$ admit a perfect matching?



The fact that we have the expansion factor of $c$ only implies, in general, that $|partial A| geq |A| + 1$, and I don't know to what extent that may be helpful. But maybe for $n$ large enough in terms of $c, k$ it may have some nice implications...










share|cite|improve this question











$endgroup$

















    0












    $begingroup$


    In chapter 1 of Lubotzky's "Discrete Groups, Expanding Graphs and Invariant Measures", remark 1.1.2, one definition of expanders is the following: if $n, k in mathbb{N}$ and $c in mathbb{R}_{> 0}$, an $(n, k, c)$-expander is a $k$-regular bipartite graph with $n$ inputs and $n$ outputs such that for any set $A$ of at most $frac{n}{2}$ inputs we have $|partial A| geq (1 + c) |A|$, where $partial A = { v : d(v, A) = 1 }$.



    The author goes on to say that any such graph admits a perfect matching by Hall's marriage lemma. Now I know that regular bipartite graphs always admit a perfect matching. However some other authors define expanders as graphs with degree bounded by $k$, instead of $k$-regular graphs. So I was wondering if the following is true:



    Let $X$ be a bipartite graph with $n$ inputs and $n$ outputs and maximum degree $k$, such that for any subset $A$ of at most $frac{n}{2}$ vertices we have $|partial A| geq (1 + c)|A|$, for some fixed $c > 0$. Does $X$ admit a perfect matching?



    The fact that we have the expansion factor of $c$ only implies, in general, that $|partial A| geq |A| + 1$, and I don't know to what extent that may be helpful. But maybe for $n$ large enough in terms of $c, k$ it may have some nice implications...










    share|cite|improve this question











    $endgroup$















      0












      0








      0





      $begingroup$


      In chapter 1 of Lubotzky's "Discrete Groups, Expanding Graphs and Invariant Measures", remark 1.1.2, one definition of expanders is the following: if $n, k in mathbb{N}$ and $c in mathbb{R}_{> 0}$, an $(n, k, c)$-expander is a $k$-regular bipartite graph with $n$ inputs and $n$ outputs such that for any set $A$ of at most $frac{n}{2}$ inputs we have $|partial A| geq (1 + c) |A|$, where $partial A = { v : d(v, A) = 1 }$.



      The author goes on to say that any such graph admits a perfect matching by Hall's marriage lemma. Now I know that regular bipartite graphs always admit a perfect matching. However some other authors define expanders as graphs with degree bounded by $k$, instead of $k$-regular graphs. So I was wondering if the following is true:



      Let $X$ be a bipartite graph with $n$ inputs and $n$ outputs and maximum degree $k$, such that for any subset $A$ of at most $frac{n}{2}$ vertices we have $|partial A| geq (1 + c)|A|$, for some fixed $c > 0$. Does $X$ admit a perfect matching?



      The fact that we have the expansion factor of $c$ only implies, in general, that $|partial A| geq |A| + 1$, and I don't know to what extent that may be helpful. But maybe for $n$ large enough in terms of $c, k$ it may have some nice implications...










      share|cite|improve this question











      $endgroup$




      In chapter 1 of Lubotzky's "Discrete Groups, Expanding Graphs and Invariant Measures", remark 1.1.2, one definition of expanders is the following: if $n, k in mathbb{N}$ and $c in mathbb{R}_{> 0}$, an $(n, k, c)$-expander is a $k$-regular bipartite graph with $n$ inputs and $n$ outputs such that for any set $A$ of at most $frac{n}{2}$ inputs we have $|partial A| geq (1 + c) |A|$, where $partial A = { v : d(v, A) = 1 }$.



      The author goes on to say that any such graph admits a perfect matching by Hall's marriage lemma. Now I know that regular bipartite graphs always admit a perfect matching. However some other authors define expanders as graphs with degree bounded by $k$, instead of $k$-regular graphs. So I was wondering if the following is true:



      Let $X$ be a bipartite graph with $n$ inputs and $n$ outputs and maximum degree $k$, such that for any subset $A$ of at most $frac{n}{2}$ vertices we have $|partial A| geq (1 + c)|A|$, for some fixed $c > 0$. Does $X$ admit a perfect matching?



      The fact that we have the expansion factor of $c$ only implies, in general, that $|partial A| geq |A| + 1$, and I don't know to what extent that may be helpful. But maybe for $n$ large enough in terms of $c, k$ it may have some nice implications...







      graph-theory bipartite-graph matching-theory






      share|cite|improve this question















      share|cite|improve this question













      share|cite|improve this question




      share|cite|improve this question








      edited Dec 11 '18 at 16:04







      frafour

















      asked Dec 11 '18 at 13:01









      frafourfrafour

      890312




      890312






















          1 Answer
          1






          active

          oldest

          votes


















          1












          $begingroup$

          Yes expansion on both sides does imply a perfect matching, if both sides of the graph have the same number of vertices--regularity is not needed.



          Let us write the sides of $G$ as $S$ and $T$; both have $n$ vertices each. Let $A$ be a subset of more than $frac{n}{2}$ vertices on one side $S$ of the graph. Then let $B= T setminus N_G(A)$.



          Now, $|N_G(A)|$ is at least $frac{n}{2}$, as subsets on one side with $frac{n}{2}$ vertices (in particular, any subset of $A$ with $frac{n}{2}$ vertices) have at least $(1+c)$ times as many neighbours on the other side. So $B$ has no more than $frac{n}{2}$ vertices and so $|N_G(B)| geq (1+c)B$. Furthemore, $N_G(B)$ does not intersect $A$.



          So writing $|A| = an$ and $|B|=bn$. Then on the one hand it follows that $|N_G(A)| = |T setminus B| = (1-b)n$. On the other hand, as $N_G(B)$ does not intersect $A$ it follows that $(1+c)bn leq n-an$. So putting these together gives $|N_G(A)| geq left(1-frac{1-a}{1+c}right)n$ $geq an = |A|$.



          So $|N_G(A)| geq |A|$ for all $A subseteq S$, and one can use the same line of reasoning to conclude that $N_G(L) geq |L|$ for all $L subseteq T$. So finish using Halls'.





          The above argument does NOT hold if one side of $G$ has more vertices than the other; i.e., if $G$ is a concentrator. In that case (wrting $|S|=n$) the equation $|N_G(A)| = (1-b)n$ does not hold if $|T|$ is less than $n$.



          ETA: And, per the comments below, there is not necessarily a perfect matching if there is expansion from only one side (and such bipartite graphs exist even when both sides have the same number $n$ vertices).






          share|cite|improve this answer











          $endgroup$













          • $begingroup$
            Thank you for your answer, everything is very clear. I just realized that in the definition of expansion in Lubotzky he demands that only the inputs have the expansion property. I imagine that in that case, there is not necessarily a perfect matching. Is that the case? If for example we assume connectedness and same number of vertices, so we cannot just use a concentrator with some isolated vertices on one side to reach the same cardinality
            $endgroup$
            – frafour
            Dec 11 '18 at 22:38






          • 1




            $begingroup$
            Ok I just found one: take a 5-3 complete bipartite graph, add two outputs and connect them to the same input. Then if a set of inputs has at most 5/2 elements, so 2 elements, then it has at least 3 neighbours. Therefore taking c = 1/2 works: we have the expanding property for small enough subsets. But we do not have a perfect matching clearly
            $endgroup$
            – frafour
            Dec 11 '18 at 22:53






          • 1




            $begingroup$
            OR you could even take a 8-6 complete bipartite graph and add two outputs and connect them to the same input; if a set of inputs has at most $5>frac{8}{2}$ elements then it has at least 6 neighbours. But yes indeed, you do need the expansion property on both sides, or rather something additional besides just expansion on one side. Fortunately the constructions in Lubotzky's book give as they use the 2nd-largest eigenvalue of the adjacency matrix to prove expansion.
            $endgroup$
            – Mike
            Dec 11 '18 at 23:08












          • $begingroup$
            We could even generalize this to make arbitrarily large counterexamples: take a 2n-(n+1) complete bipartite graph, add (n-1) vertices on the right and connect them all to the same vertex. I wonder if there are arbitrarily large examples of bounded degree, and/or for a fixed constant c, or if for n large enough it eventually works...
            $endgroup$
            – frafour
            Dec 11 '18 at 23:12








          • 1




            $begingroup$
            Well, you could come up with a bounded-degree concentrator w $n$ inputs, say $3n/4$ outputs s.t subsets of the input w fewer than $frac{n}{2}$ vertices have as many neighbours in the output [these such constructions exist--even explicit constructions for arbitrarily large $n$--the degree would be large but bounded]. Then add $n/4$ additional outputs and attach every (say) 4 of these added outputs to the same input [so each additional output has degreee 1 and each vertex in the input has degree at most 4 from what it had before]
            $endgroup$
            – Mike
            Dec 11 '18 at 23:20














          Your Answer





          StackExchange.ifUsing("editor", function () {
          return StackExchange.using("mathjaxEditing", function () {
          StackExchange.MarkdownEditor.creationCallbacks.add(function (editor, postfix) {
          StackExchange.mathjaxEditing.prepareWmdForMathJax(editor, postfix, [["$", "$"], ["\\(","\\)"]]);
          });
          });
          }, "mathjax-editing");

          StackExchange.ready(function() {
          var channelOptions = {
          tags: "".split(" "),
          id: "69"
          };
          initTagRenderer("".split(" "), "".split(" "), channelOptions);

          StackExchange.using("externalEditor", function() {
          // Have to fire editor after snippets, if snippets enabled
          if (StackExchange.settings.snippets.snippetsEnabled) {
          StackExchange.using("snippets", function() {
          createEditor();
          });
          }
          else {
          createEditor();
          }
          });

          function createEditor() {
          StackExchange.prepareEditor({
          heartbeatType: 'answer',
          autoActivateHeartbeat: false,
          convertImagesToLinks: true,
          noModals: true,
          showLowRepImageUploadWarning: true,
          reputationToPostImages: 10,
          bindNavPrevention: true,
          postfix: "",
          imageUploader: {
          brandingHtml: "Powered by u003ca class="icon-imgur-white" href="https://imgur.com/"u003eu003c/au003e",
          contentPolicyHtml: "User contributions licensed under u003ca href="https://creativecommons.org/licenses/by-sa/3.0/"u003ecc by-sa 3.0 with attribution requiredu003c/au003e u003ca href="https://stackoverflow.com/legal/content-policy"u003e(content policy)u003c/au003e",
          allowUrls: true
          },
          noCode: true, onDemand: true,
          discardSelector: ".discard-answer"
          ,immediatelyShowMarkdownHelp:true
          });


          }
          });














          draft saved

          draft discarded


















          StackExchange.ready(
          function () {
          StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f3035264%2fweaker-version-of-halls-condition%23new-answer', 'question_page');
          }
          );

          Post as a guest















          Required, but never shown

























          1 Answer
          1






          active

          oldest

          votes








          1 Answer
          1






          active

          oldest

          votes









          active

          oldest

          votes






          active

          oldest

          votes









          1












          $begingroup$

          Yes expansion on both sides does imply a perfect matching, if both sides of the graph have the same number of vertices--regularity is not needed.



          Let us write the sides of $G$ as $S$ and $T$; both have $n$ vertices each. Let $A$ be a subset of more than $frac{n}{2}$ vertices on one side $S$ of the graph. Then let $B= T setminus N_G(A)$.



          Now, $|N_G(A)|$ is at least $frac{n}{2}$, as subsets on one side with $frac{n}{2}$ vertices (in particular, any subset of $A$ with $frac{n}{2}$ vertices) have at least $(1+c)$ times as many neighbours on the other side. So $B$ has no more than $frac{n}{2}$ vertices and so $|N_G(B)| geq (1+c)B$. Furthemore, $N_G(B)$ does not intersect $A$.



          So writing $|A| = an$ and $|B|=bn$. Then on the one hand it follows that $|N_G(A)| = |T setminus B| = (1-b)n$. On the other hand, as $N_G(B)$ does not intersect $A$ it follows that $(1+c)bn leq n-an$. So putting these together gives $|N_G(A)| geq left(1-frac{1-a}{1+c}right)n$ $geq an = |A|$.



          So $|N_G(A)| geq |A|$ for all $A subseteq S$, and one can use the same line of reasoning to conclude that $N_G(L) geq |L|$ for all $L subseteq T$. So finish using Halls'.





          The above argument does NOT hold if one side of $G$ has more vertices than the other; i.e., if $G$ is a concentrator. In that case (wrting $|S|=n$) the equation $|N_G(A)| = (1-b)n$ does not hold if $|T|$ is less than $n$.



          ETA: And, per the comments below, there is not necessarily a perfect matching if there is expansion from only one side (and such bipartite graphs exist even when both sides have the same number $n$ vertices).






          share|cite|improve this answer











          $endgroup$













          • $begingroup$
            Thank you for your answer, everything is very clear. I just realized that in the definition of expansion in Lubotzky he demands that only the inputs have the expansion property. I imagine that in that case, there is not necessarily a perfect matching. Is that the case? If for example we assume connectedness and same number of vertices, so we cannot just use a concentrator with some isolated vertices on one side to reach the same cardinality
            $endgroup$
            – frafour
            Dec 11 '18 at 22:38






          • 1




            $begingroup$
            Ok I just found one: take a 5-3 complete bipartite graph, add two outputs and connect them to the same input. Then if a set of inputs has at most 5/2 elements, so 2 elements, then it has at least 3 neighbours. Therefore taking c = 1/2 works: we have the expanding property for small enough subsets. But we do not have a perfect matching clearly
            $endgroup$
            – frafour
            Dec 11 '18 at 22:53






          • 1




            $begingroup$
            OR you could even take a 8-6 complete bipartite graph and add two outputs and connect them to the same input; if a set of inputs has at most $5>frac{8}{2}$ elements then it has at least 6 neighbours. But yes indeed, you do need the expansion property on both sides, or rather something additional besides just expansion on one side. Fortunately the constructions in Lubotzky's book give as they use the 2nd-largest eigenvalue of the adjacency matrix to prove expansion.
            $endgroup$
            – Mike
            Dec 11 '18 at 23:08












          • $begingroup$
            We could even generalize this to make arbitrarily large counterexamples: take a 2n-(n+1) complete bipartite graph, add (n-1) vertices on the right and connect them all to the same vertex. I wonder if there are arbitrarily large examples of bounded degree, and/or for a fixed constant c, or if for n large enough it eventually works...
            $endgroup$
            – frafour
            Dec 11 '18 at 23:12








          • 1




            $begingroup$
            Well, you could come up with a bounded-degree concentrator w $n$ inputs, say $3n/4$ outputs s.t subsets of the input w fewer than $frac{n}{2}$ vertices have as many neighbours in the output [these such constructions exist--even explicit constructions for arbitrarily large $n$--the degree would be large but bounded]. Then add $n/4$ additional outputs and attach every (say) 4 of these added outputs to the same input [so each additional output has degreee 1 and each vertex in the input has degree at most 4 from what it had before]
            $endgroup$
            – Mike
            Dec 11 '18 at 23:20


















          1












          $begingroup$

          Yes expansion on both sides does imply a perfect matching, if both sides of the graph have the same number of vertices--regularity is not needed.



          Let us write the sides of $G$ as $S$ and $T$; both have $n$ vertices each. Let $A$ be a subset of more than $frac{n}{2}$ vertices on one side $S$ of the graph. Then let $B= T setminus N_G(A)$.



          Now, $|N_G(A)|$ is at least $frac{n}{2}$, as subsets on one side with $frac{n}{2}$ vertices (in particular, any subset of $A$ with $frac{n}{2}$ vertices) have at least $(1+c)$ times as many neighbours on the other side. So $B$ has no more than $frac{n}{2}$ vertices and so $|N_G(B)| geq (1+c)B$. Furthemore, $N_G(B)$ does not intersect $A$.



          So writing $|A| = an$ and $|B|=bn$. Then on the one hand it follows that $|N_G(A)| = |T setminus B| = (1-b)n$. On the other hand, as $N_G(B)$ does not intersect $A$ it follows that $(1+c)bn leq n-an$. So putting these together gives $|N_G(A)| geq left(1-frac{1-a}{1+c}right)n$ $geq an = |A|$.



          So $|N_G(A)| geq |A|$ for all $A subseteq S$, and one can use the same line of reasoning to conclude that $N_G(L) geq |L|$ for all $L subseteq T$. So finish using Halls'.





          The above argument does NOT hold if one side of $G$ has more vertices than the other; i.e., if $G$ is a concentrator. In that case (wrting $|S|=n$) the equation $|N_G(A)| = (1-b)n$ does not hold if $|T|$ is less than $n$.



          ETA: And, per the comments below, there is not necessarily a perfect matching if there is expansion from only one side (and such bipartite graphs exist even when both sides have the same number $n$ vertices).






          share|cite|improve this answer











          $endgroup$













          • $begingroup$
            Thank you for your answer, everything is very clear. I just realized that in the definition of expansion in Lubotzky he demands that only the inputs have the expansion property. I imagine that in that case, there is not necessarily a perfect matching. Is that the case? If for example we assume connectedness and same number of vertices, so we cannot just use a concentrator with some isolated vertices on one side to reach the same cardinality
            $endgroup$
            – frafour
            Dec 11 '18 at 22:38






          • 1




            $begingroup$
            Ok I just found one: take a 5-3 complete bipartite graph, add two outputs and connect them to the same input. Then if a set of inputs has at most 5/2 elements, so 2 elements, then it has at least 3 neighbours. Therefore taking c = 1/2 works: we have the expanding property for small enough subsets. But we do not have a perfect matching clearly
            $endgroup$
            – frafour
            Dec 11 '18 at 22:53






          • 1




            $begingroup$
            OR you could even take a 8-6 complete bipartite graph and add two outputs and connect them to the same input; if a set of inputs has at most $5>frac{8}{2}$ elements then it has at least 6 neighbours. But yes indeed, you do need the expansion property on both sides, or rather something additional besides just expansion on one side. Fortunately the constructions in Lubotzky's book give as they use the 2nd-largest eigenvalue of the adjacency matrix to prove expansion.
            $endgroup$
            – Mike
            Dec 11 '18 at 23:08












          • $begingroup$
            We could even generalize this to make arbitrarily large counterexamples: take a 2n-(n+1) complete bipartite graph, add (n-1) vertices on the right and connect them all to the same vertex. I wonder if there are arbitrarily large examples of bounded degree, and/or for a fixed constant c, or if for n large enough it eventually works...
            $endgroup$
            – frafour
            Dec 11 '18 at 23:12








          • 1




            $begingroup$
            Well, you could come up with a bounded-degree concentrator w $n$ inputs, say $3n/4$ outputs s.t subsets of the input w fewer than $frac{n}{2}$ vertices have as many neighbours in the output [these such constructions exist--even explicit constructions for arbitrarily large $n$--the degree would be large but bounded]. Then add $n/4$ additional outputs and attach every (say) 4 of these added outputs to the same input [so each additional output has degreee 1 and each vertex in the input has degree at most 4 from what it had before]
            $endgroup$
            – Mike
            Dec 11 '18 at 23:20
















          1












          1








          1





          $begingroup$

          Yes expansion on both sides does imply a perfect matching, if both sides of the graph have the same number of vertices--regularity is not needed.



          Let us write the sides of $G$ as $S$ and $T$; both have $n$ vertices each. Let $A$ be a subset of more than $frac{n}{2}$ vertices on one side $S$ of the graph. Then let $B= T setminus N_G(A)$.



          Now, $|N_G(A)|$ is at least $frac{n}{2}$, as subsets on one side with $frac{n}{2}$ vertices (in particular, any subset of $A$ with $frac{n}{2}$ vertices) have at least $(1+c)$ times as many neighbours on the other side. So $B$ has no more than $frac{n}{2}$ vertices and so $|N_G(B)| geq (1+c)B$. Furthemore, $N_G(B)$ does not intersect $A$.



          So writing $|A| = an$ and $|B|=bn$. Then on the one hand it follows that $|N_G(A)| = |T setminus B| = (1-b)n$. On the other hand, as $N_G(B)$ does not intersect $A$ it follows that $(1+c)bn leq n-an$. So putting these together gives $|N_G(A)| geq left(1-frac{1-a}{1+c}right)n$ $geq an = |A|$.



          So $|N_G(A)| geq |A|$ for all $A subseteq S$, and one can use the same line of reasoning to conclude that $N_G(L) geq |L|$ for all $L subseteq T$. So finish using Halls'.





          The above argument does NOT hold if one side of $G$ has more vertices than the other; i.e., if $G$ is a concentrator. In that case (wrting $|S|=n$) the equation $|N_G(A)| = (1-b)n$ does not hold if $|T|$ is less than $n$.



          ETA: And, per the comments below, there is not necessarily a perfect matching if there is expansion from only one side (and such bipartite graphs exist even when both sides have the same number $n$ vertices).






          share|cite|improve this answer











          $endgroup$



          Yes expansion on both sides does imply a perfect matching, if both sides of the graph have the same number of vertices--regularity is not needed.



          Let us write the sides of $G$ as $S$ and $T$; both have $n$ vertices each. Let $A$ be a subset of more than $frac{n}{2}$ vertices on one side $S$ of the graph. Then let $B= T setminus N_G(A)$.



          Now, $|N_G(A)|$ is at least $frac{n}{2}$, as subsets on one side with $frac{n}{2}$ vertices (in particular, any subset of $A$ with $frac{n}{2}$ vertices) have at least $(1+c)$ times as many neighbours on the other side. So $B$ has no more than $frac{n}{2}$ vertices and so $|N_G(B)| geq (1+c)B$. Furthemore, $N_G(B)$ does not intersect $A$.



          So writing $|A| = an$ and $|B|=bn$. Then on the one hand it follows that $|N_G(A)| = |T setminus B| = (1-b)n$. On the other hand, as $N_G(B)$ does not intersect $A$ it follows that $(1+c)bn leq n-an$. So putting these together gives $|N_G(A)| geq left(1-frac{1-a}{1+c}right)n$ $geq an = |A|$.



          So $|N_G(A)| geq |A|$ for all $A subseteq S$, and one can use the same line of reasoning to conclude that $N_G(L) geq |L|$ for all $L subseteq T$. So finish using Halls'.





          The above argument does NOT hold if one side of $G$ has more vertices than the other; i.e., if $G$ is a concentrator. In that case (wrting $|S|=n$) the equation $|N_G(A)| = (1-b)n$ does not hold if $|T|$ is less than $n$.



          ETA: And, per the comments below, there is not necessarily a perfect matching if there is expansion from only one side (and such bipartite graphs exist even when both sides have the same number $n$ vertices).







          share|cite|improve this answer














          share|cite|improve this answer



          share|cite|improve this answer








          edited Dec 12 '18 at 16:54

























          answered Dec 11 '18 at 21:29









          MikeMike

          4,536512




          4,536512












          • $begingroup$
            Thank you for your answer, everything is very clear. I just realized that in the definition of expansion in Lubotzky he demands that only the inputs have the expansion property. I imagine that in that case, there is not necessarily a perfect matching. Is that the case? If for example we assume connectedness and same number of vertices, so we cannot just use a concentrator with some isolated vertices on one side to reach the same cardinality
            $endgroup$
            – frafour
            Dec 11 '18 at 22:38






          • 1




            $begingroup$
            Ok I just found one: take a 5-3 complete bipartite graph, add two outputs and connect them to the same input. Then if a set of inputs has at most 5/2 elements, so 2 elements, then it has at least 3 neighbours. Therefore taking c = 1/2 works: we have the expanding property for small enough subsets. But we do not have a perfect matching clearly
            $endgroup$
            – frafour
            Dec 11 '18 at 22:53






          • 1




            $begingroup$
            OR you could even take a 8-6 complete bipartite graph and add two outputs and connect them to the same input; if a set of inputs has at most $5>frac{8}{2}$ elements then it has at least 6 neighbours. But yes indeed, you do need the expansion property on both sides, or rather something additional besides just expansion on one side. Fortunately the constructions in Lubotzky's book give as they use the 2nd-largest eigenvalue of the adjacency matrix to prove expansion.
            $endgroup$
            – Mike
            Dec 11 '18 at 23:08












          • $begingroup$
            We could even generalize this to make arbitrarily large counterexamples: take a 2n-(n+1) complete bipartite graph, add (n-1) vertices on the right and connect them all to the same vertex. I wonder if there are arbitrarily large examples of bounded degree, and/or for a fixed constant c, or if for n large enough it eventually works...
            $endgroup$
            – frafour
            Dec 11 '18 at 23:12








          • 1




            $begingroup$
            Well, you could come up with a bounded-degree concentrator w $n$ inputs, say $3n/4$ outputs s.t subsets of the input w fewer than $frac{n}{2}$ vertices have as many neighbours in the output [these such constructions exist--even explicit constructions for arbitrarily large $n$--the degree would be large but bounded]. Then add $n/4$ additional outputs and attach every (say) 4 of these added outputs to the same input [so each additional output has degreee 1 and each vertex in the input has degree at most 4 from what it had before]
            $endgroup$
            – Mike
            Dec 11 '18 at 23:20




















          • $begingroup$
            Thank you for your answer, everything is very clear. I just realized that in the definition of expansion in Lubotzky he demands that only the inputs have the expansion property. I imagine that in that case, there is not necessarily a perfect matching. Is that the case? If for example we assume connectedness and same number of vertices, so we cannot just use a concentrator with some isolated vertices on one side to reach the same cardinality
            $endgroup$
            – frafour
            Dec 11 '18 at 22:38






          • 1




            $begingroup$
            Ok I just found one: take a 5-3 complete bipartite graph, add two outputs and connect them to the same input. Then if a set of inputs has at most 5/2 elements, so 2 elements, then it has at least 3 neighbours. Therefore taking c = 1/2 works: we have the expanding property for small enough subsets. But we do not have a perfect matching clearly
            $endgroup$
            – frafour
            Dec 11 '18 at 22:53






          • 1




            $begingroup$
            OR you could even take a 8-6 complete bipartite graph and add two outputs and connect them to the same input; if a set of inputs has at most $5>frac{8}{2}$ elements then it has at least 6 neighbours. But yes indeed, you do need the expansion property on both sides, or rather something additional besides just expansion on one side. Fortunately the constructions in Lubotzky's book give as they use the 2nd-largest eigenvalue of the adjacency matrix to prove expansion.
            $endgroup$
            – Mike
            Dec 11 '18 at 23:08












          • $begingroup$
            We could even generalize this to make arbitrarily large counterexamples: take a 2n-(n+1) complete bipartite graph, add (n-1) vertices on the right and connect them all to the same vertex. I wonder if there are arbitrarily large examples of bounded degree, and/or for a fixed constant c, or if for n large enough it eventually works...
            $endgroup$
            – frafour
            Dec 11 '18 at 23:12








          • 1




            $begingroup$
            Well, you could come up with a bounded-degree concentrator w $n$ inputs, say $3n/4$ outputs s.t subsets of the input w fewer than $frac{n}{2}$ vertices have as many neighbours in the output [these such constructions exist--even explicit constructions for arbitrarily large $n$--the degree would be large but bounded]. Then add $n/4$ additional outputs and attach every (say) 4 of these added outputs to the same input [so each additional output has degreee 1 and each vertex in the input has degree at most 4 from what it had before]
            $endgroup$
            – Mike
            Dec 11 '18 at 23:20


















          $begingroup$
          Thank you for your answer, everything is very clear. I just realized that in the definition of expansion in Lubotzky he demands that only the inputs have the expansion property. I imagine that in that case, there is not necessarily a perfect matching. Is that the case? If for example we assume connectedness and same number of vertices, so we cannot just use a concentrator with some isolated vertices on one side to reach the same cardinality
          $endgroup$
          – frafour
          Dec 11 '18 at 22:38




          $begingroup$
          Thank you for your answer, everything is very clear. I just realized that in the definition of expansion in Lubotzky he demands that only the inputs have the expansion property. I imagine that in that case, there is not necessarily a perfect matching. Is that the case? If for example we assume connectedness and same number of vertices, so we cannot just use a concentrator with some isolated vertices on one side to reach the same cardinality
          $endgroup$
          – frafour
          Dec 11 '18 at 22:38




          1




          1




          $begingroup$
          Ok I just found one: take a 5-3 complete bipartite graph, add two outputs and connect them to the same input. Then if a set of inputs has at most 5/2 elements, so 2 elements, then it has at least 3 neighbours. Therefore taking c = 1/2 works: we have the expanding property for small enough subsets. But we do not have a perfect matching clearly
          $endgroup$
          – frafour
          Dec 11 '18 at 22:53




          $begingroup$
          Ok I just found one: take a 5-3 complete bipartite graph, add two outputs and connect them to the same input. Then if a set of inputs has at most 5/2 elements, so 2 elements, then it has at least 3 neighbours. Therefore taking c = 1/2 works: we have the expanding property for small enough subsets. But we do not have a perfect matching clearly
          $endgroup$
          – frafour
          Dec 11 '18 at 22:53




          1




          1




          $begingroup$
          OR you could even take a 8-6 complete bipartite graph and add two outputs and connect them to the same input; if a set of inputs has at most $5>frac{8}{2}$ elements then it has at least 6 neighbours. But yes indeed, you do need the expansion property on both sides, or rather something additional besides just expansion on one side. Fortunately the constructions in Lubotzky's book give as they use the 2nd-largest eigenvalue of the adjacency matrix to prove expansion.
          $endgroup$
          – Mike
          Dec 11 '18 at 23:08






          $begingroup$
          OR you could even take a 8-6 complete bipartite graph and add two outputs and connect them to the same input; if a set of inputs has at most $5>frac{8}{2}$ elements then it has at least 6 neighbours. But yes indeed, you do need the expansion property on both sides, or rather something additional besides just expansion on one side. Fortunately the constructions in Lubotzky's book give as they use the 2nd-largest eigenvalue of the adjacency matrix to prove expansion.
          $endgroup$
          – Mike
          Dec 11 '18 at 23:08














          $begingroup$
          We could even generalize this to make arbitrarily large counterexamples: take a 2n-(n+1) complete bipartite graph, add (n-1) vertices on the right and connect them all to the same vertex. I wonder if there are arbitrarily large examples of bounded degree, and/or for a fixed constant c, or if for n large enough it eventually works...
          $endgroup$
          – frafour
          Dec 11 '18 at 23:12






          $begingroup$
          We could even generalize this to make arbitrarily large counterexamples: take a 2n-(n+1) complete bipartite graph, add (n-1) vertices on the right and connect them all to the same vertex. I wonder if there are arbitrarily large examples of bounded degree, and/or for a fixed constant c, or if for n large enough it eventually works...
          $endgroup$
          – frafour
          Dec 11 '18 at 23:12






          1




          1




          $begingroup$
          Well, you could come up with a bounded-degree concentrator w $n$ inputs, say $3n/4$ outputs s.t subsets of the input w fewer than $frac{n}{2}$ vertices have as many neighbours in the output [these such constructions exist--even explicit constructions for arbitrarily large $n$--the degree would be large but bounded]. Then add $n/4$ additional outputs and attach every (say) 4 of these added outputs to the same input [so each additional output has degreee 1 and each vertex in the input has degree at most 4 from what it had before]
          $endgroup$
          – Mike
          Dec 11 '18 at 23:20






          $begingroup$
          Well, you could come up with a bounded-degree concentrator w $n$ inputs, say $3n/4$ outputs s.t subsets of the input w fewer than $frac{n}{2}$ vertices have as many neighbours in the output [these such constructions exist--even explicit constructions for arbitrarily large $n$--the degree would be large but bounded]. Then add $n/4$ additional outputs and attach every (say) 4 of these added outputs to the same input [so each additional output has degreee 1 and each vertex in the input has degree at most 4 from what it had before]
          $endgroup$
          – Mike
          Dec 11 '18 at 23:20




















          draft saved

          draft discarded




















































          Thanks for contributing an answer to Mathematics Stack Exchange!


          • Please be sure to answer the question. Provide details and share your research!

          But avoid



          • Asking for help, clarification, or responding to other answers.

          • Making statements based on opinion; back them up with references or personal experience.


          Use MathJax to format equations. MathJax reference.


          To learn more, see our tips on writing great answers.




          draft saved


          draft discarded














          StackExchange.ready(
          function () {
          StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fmath.stackexchange.com%2fquestions%2f3035264%2fweaker-version-of-halls-condition%23new-answer', 'question_page');
          }
          );

          Post as a guest















          Required, but never shown





















































          Required, but never shown














          Required, but never shown












          Required, but never shown







          Required, but never shown

































          Required, but never shown














          Required, but never shown












          Required, but never shown







          Required, but never shown







          Popular posts from this blog

          Biblatex bibliography style without URLs when DOI exists (in Overleaf with Zotero bibliography)

          ComboBox Display Member on multiple fields

          Is it possible to collect Nectar points via Trainline?